Friday, March 22, 2013

Can you identify the plane that just flew by?

Apple's Siri has surprised everyone with its abilities to understand human voice and take action on various commands. It has brought human interaction to the digital world closer than ever before. But what if I tell you that my application Stri (Sixth Sense Search, pronounced as s-T-ree) can understand and answer humans like humans?
Lets try an example in google.com. Search for "Can you identify the plane that just flew by?". Here is the result.

Ooops. No result. Google couldnt tell you that it didnt know the answer. What it could do is to search for pages that has the maximum hits to the keywords you wrote! Are you satisfied?
Lets see how Stri handles the same request. Here is the result. 



So what are we doing here?
We, in Stri, have revolutionized the way digital data is interpreted, stored and retrieved. When a user searches for content, Stri analyses the content and looks for real life data that matches the query. The most difficult aspect is to understand the different objects in the query, use different real-time parameters such as location and time to build the appropriate question. For this specific example, we have identified the search query as "Can you identify the plane that flew by at 12.30PM in Sanjaynagar Bangalore?". The next is to identify on what is the search query related. Here the object is "Plane". So we have 3 critical information with us: Plane, 12.30PM and Sanjaynagar. Finally, the action is to be done is "Identify". Multiple keywords would still mean the same, such as, list, find, identify, search etc. Finally, we have identified the task: "Search for planes that flew over Sanjaynagar at 12.30PM"!!
Stri's approach to solve the problem is to use open databases. Stri has provided a wrapper to many services such as flight, train, public transport, media, tv etc. With the wrappers, Stri can connect and fetch details for various query string, retrieve results and post it back to the Stri search engine. Stri engine then collects information, with confidence numbers and posts the results.
The technology can now be extended to a number of dimensions! One can actually search when he visited Paris based on flight details / email details or check-in information in facebook. Or it can extend to devices that the user owns or his friends. This is definitely a wave that will carry everyone with it on the internet!



3 comments:

  1. The contrast between the two images of Google and Stri search give an excellent peek into what we are getting into. It would be much smarter than what Google provides today. What would be nice to understand is what are the key challenges in achieving this? Are we almost there? Or are there any bottlenecks?

    Hope you get to pursue this thread. All the best.

    Vinay

    ReplyDelete
  2. Thanks so much Vinay! I need a lot of time to research and find out where others are with respect to this technology.
    However, the idea of offloading this to research people makes sense :)
    Regards,
    Varchas R S

    ReplyDelete
  3. Varchas, Pl see if this paper titled "I can sense it" produced from IIT Bombay gives any direction. At least the title looks like a good response to your title :-) A good place to start is to look for an open source system where a bunch of such algorithms are implemented and then play with it. I don't know if IITB has one.

    ReplyDelete